The use of marker-assisted selection (MAS) to predict genetic value of breeding lines is increasing in private and public plant breeding. MAS is an attractive alternative to phenotypic selection because MAS can be performed on a single plant or seed and decrease selection cycle duration. Advancements in genotyping are rapidly decreasing marker costs so that genotyping is becoming cheaper than phenotyping. Thus, the potential of MAS to achieve greater gains from selection per unit time and cost than phenotypic selection is growing. The ability to achieve genome-wide genotyping, however, may not be best utilized by conventional-MAS methods that have proven to be largely ineffective for improving the complex quantitative traits that dictate the success of new crop varieties. An emerging alternative to MAS is a technique termed genomic selection (GS) that uses a random-effects statistical modeling approach to jointly estimate all marker effects. This method does not require significance testing and has the goal of capturing small-effect QTL that are excluded by significance thresholds used in conventionalMAS. The use of GS is becoming a popular tool in animal breeding and is garnering the attention of plant breeders; however, evidence regarding the performance and the best methodology for applying GS in plant breeding is currently limited. In this research, GS was compared to conventional-MAS and phenotypic selection (PS) by deterministic simulation and empirical evaluations in plant breeding. Performance of these methods was empirically tested in two biparental wheat populations and in an advanced wheat breeding population comprised of multiple families derived from many different crosses. These studies showed that GS was superior to conventional-MAS in predicting the genetic value of breeding lines and that GS was competitive with PS in terms of accuracy. Furthermore, results indicate that GS could significantly reduce the selection cycle duration and achieve prediction accuracies that would enable plant breeders to achieve greater gains per unit time and cost than are possible with current MAS strategies.